Risk of prostate and bladder cancers in patients with spinal cord injury: A population-based cohort study.

Risk

ofprostateandbladdercancersinpatientswithspinalcordinjury : A population-basedcohortstudy

Wen-Yuan Lee,M.D.

, Li-

MinSun,M.D.

, Cheng-LiLin,MSc

, Ji-AnLiang,M.D.

, Yen-Jung Chang,Ph.D

, Fung-ChangSung,Ph.D

, Chia-HungKao,M.D.

1. Introduction Spinalcordinjury(SCI)hasareportedincidenceinNorth Americaof15to55peoplepermillion [1–3]. Theincidence in

Taiwanisconsiderablyhigherat150peoplepermillion, accordingtonationwidedata [4].

Advancesintechnologyfor thetreatmentandfollow-upcareofpeoplewithSCIhave signifcantlyimprovedthelifeexpectancyofthesepatientsto

fguresapproachingthatofthegeneralpopulation [5]. Thus, mostpatientswithSCIarelikelytoencounterproblems

associatedwithaging,includingcancer [6]. Prostate canceristhe ffth

mostcommoncanceramong Taiwanese men.Theage-adjustedincidenceofprostate cancer amongTaiwanesemenin2008was24.55per 100,000 [7].

Earlierstudieshadreportedalowerincidence of

prostatecanceramongpatientswithSCIcomparedwith the generalpopulation [8–12].

Bycontrast,patientswith SCI areknowntobeatanincreasedriskofbladdercancer compared withthegeneralpopulation,asshowninseveral studies [13–16].

Onestudyshowedthatahigherspinal level ofparalysiswasassociatedwithalowerriskof prostate cancer [11]. Separatestudiessuggestedthatearlier onset orlongerduration,orboth,ofSCImightbe associated withalowerriskofprostatecancer [8,9]. We

investigatedthesepossibleassociationsthroughanalysesto evaluate whetherourdatacompliedwiththe fndings of previous reports.Nolargepopulation-basedstudieshad evaluated

theassociationsbetweenbladderandprostate cancer andSCI;therefore,weinvestigatedwhetherTaiwa- nese patientswithSCIshowedthesamepatternsofcancer risk asthatreportedinWesterncountries. 2.

Materialsandmethods 2.1. Datasources Data wereaccessedfromtheNationalHealthInsurance Research Database(NHIRD)administeredbytheTaiwan National HealthResearchInstitute.Thisstudywas approved bytheEthicsReviewBoardoftheChinaMedical University (CMU-REC-101-

012).Thisinsuranceprogram was initiatedin1996andcoversapproximately99%ofthe 23.74

millionpeoplelivinginTaiwan [17]. TheInterna- tional Classifcations ofDisease,ninthRevision,Clinical Modifcation (ICD-9-CM)codeswereusedtodefne dis- eases intheclaimsdata. 2.2.

Studyparticipants From theclaimsdataforinpatientsfrom1997to2010, patients

withnewlydiagnosedSCI(ICD-9-CMcodes806 and 952)wereselectedastheSCIcohort.Theprimary purpose ofthisstudywastoexaminetheassociation between SCIandcancer.Hence,theexclusioncriteriawere patients youngerthan20yearsandthosewithahistoryof any

canceratbaseline.ThedateofdiagnosisofSCI(as recorded intheinpatientclaimsdata)wasdefned asthe index date.ForeachSCIcase,werandomlyselected4 people withoutSCIfromtheNHIbenefciaries toforma comparison groupasthegeneralpopulation.Thesepatients were

alsorequiredtobe20yearsorolder;thecomparison participants

werefrequencymatchedtothepatientsbysex, age (5-yinterval),andindexdate.Theconfrmationof bladder cancerorprostatecancereventswasbasedonthe registry ofcatastrophicillness,whichwasasubsetofthe NHIRD data.Histologicalorpathologicalconfrmation or both

foreachcancerwererequiredforeachcase.Therich automatic

innervationoftheprostateglandentersthespinal cord atT10orhigher [18], andinterruptionofthesetracts by completecordlesionsaboveT11interfereswith hormone secretion.AccordingtotheICDcode,various levels ofSCIisfurtherdividedintoC1-C4,C5-C7,T1-T6, T7-T12,

lumbar,sacral,andcaudaequinaSCI,sowewere unable toidentifytheexactlocationofSCIaboveorbelow T11. Instead,wedividedourpatientwithSCIatthelevelof T6-T7,

becauseitisalmostthemiddleofthewholespine, with acomparablenumberineachgroup. 2.3.

Outcomemeasures Each patientwasfollowedfromtheindexdateuntilthe occurrence

ofoneofthefollowingevents:bladdercancer diagnosis (ICD-9-CM188),prostatecancerdiagnosis(ICD- 9- CM 185),death,withdrawalfromtheNHI,oruntil December 31,2010,withestimatedfollow-upperson- years. Data oncomorbiditieswerealsoanalyzed,whichincluded hypertension (ICD-9-CM401- 405),diabetes(ICD-9-CM 250), andhyperlipidemia(ICD-9-CM272).Comorbidities were

requiredtohavebeendiagnosedbeforetheindexdate. 2.4. Statisticalanalysis Distributions

ofage(r49, 50–64, and65þ y), sex,and comorbidity werecomparedbetweentheSCIandnon-SCI cohorts.

Thedifferencesbetweenthegroupsweretested using thechi-squaretest.Thesex-,age-,andcomorbidity- specifc incidencedensities(per10,000person-years)of bladder

cancerandprostatecancerweremeasuredineach cohort. TheKaplan-MeierapproachandstandardCox regression analysismayleadtobiasedestimationsof cumulative

incidenceandeffectsofcovariatesondisease risks [19]. WeusedtheFineandGraymodel [20], which extends thestandardCoxproportionalhazardmodelto estimate

thecumulativeincidenceofthebladdercancerand prostate

cancer,accountingforthecompetingriskofdeaths. Deaths withoutbladdercancerandprostatecancerwere identifed fromtheinpatientclaims.Thecompetingrisk analysis andtheKaplan-Meiercumulativeincidence fgures were performedandplottedusingSTATASE11(Stata Corp,

CollegeStation,TX).Datamanagementanddescrip- tive analyseswereperformedusingtheSAS9.2statistical package (SASInstituteInc.,Cary,NC). 3. Results Our

datasetincluded272,005participantsacrossboththe patient

andcomparisongroups.Themeanagewasstatisti- cally similarinbothgroups.The54,401patientswithSCI were classifed accordingtothespinalleveloftheirinjury as

follows:29,565cervicalspine,8,308thoracicspine, 16,451 lumbar-sacral-

coccygealspine,and77patientswith unclassifed SCI.TheSCIcohortcontainedahigher proportion

ofmen(62.8%)thanwomen,andpatientswere relatively young(45.8%aged49yoryounger).Compared with thenon-SCIcohort,theSCIcohortshowedahigher percentage

ofdiabetes,hypertension,andhyperlipidemia (P o 0.0001) (Table 1). The

overallincidenceofbladdercancerwas12%lower in theSCIcohortthanthenon-SCIcohort(2.56vs.2.82per 10,000 person-years)withanadjustedhazardratio(HR)of 0.91 (95%confdence interval[CI] ¼ 0.72–1.16) (Table 2). The incidencerateratio(IRR)ofbladdercancerwashigher for womenthanmen(IRR ¼ 1.10, 95%CI ¼ 0.72–1.68 vs. IRR ¼ 0.80, 95%CI ¼ 0.61–1.07). However,the adjusted

HRwasnotstatisticallyinsignifcant forwomen and men.TheIRRofbladdercancerwashighestamong younger people(r49 yofage,IRR ¼ 1.52, 95%CI ¼ 0.80–2.88) andlowestamongolderpeople(65þ y ofage, IRR ¼ 0.82, 95%CI ¼ 0.60–1.11). However,theadjusted HR

didnotshowtheagedifference.Thecomorbiditydid not havesignifcant effectontherelationshipbetweenSCI and bladdercanceraswell. The

overallincidenceofprostatecancerwas33%lower in theSCIthannon-SCIcohort(5.29vs.7.56per10,000 person-years), withanadjustedHRof0.73(95%CI ¼ 0.59–0.90).

Eachagegroupconsistentlyhadasignifcantly lower IRRofprostatecancerforpatientswithSCI,andthe adjusted HRshowedasignifcant differencein65þ years of

agegroup.PatientswithSCIwithcomorbidityshowed a signifcantly lowerriskforprostatecancerthanthe comorbid non-SCIcohort. Table 3 shows theincidence,crude,andadjustedHRs for

bladdercancerandprostatecancerassociatedwith various levelsofSCI.Wedefned higherspinallevelof SCI asaninjurylocatedatoraboveT6,andlowerspinal level asSCIatorbelowT7.Patientswithhigher-levelSCI showed a15%lowerriskofbladdercancercompared with thecomparisongroup,butthedifferencewasnot signifcant (adjustedHR ¼ 0.85, 95%CI ¼ 0.62–1.18). By contrast,patientswithhigher-

levelSCIshoweda statisticallysignifcant 29%lowerriskforprostatecancer than thecomparisongroup(adjustedHR ¼ 0.71, 95% CI ¼ 0.54–0.93). Within the frst

5yearsafterSCIdiagnosis,theincidence of prostatecancerwaslowerintheSCIthannon-SCIcohort (6.22 vs.6.71per10,000person-years),withanadjusted HR of0.79(95%CI ¼ 0.60–1.03) (Table 4). Theriskof occurrence ofprostatecancercontinuedtodecreaseafterthe initial

5years,butthedifferencewasstillnotstatistically signifcant (adjustedHR:0.73,95%CI ¼ 0.53–1.03). Figs.

1 and 2 displaytheKaplan-Meierincidencecurves accountingforcompetingriskforthebladdercancerand prostate cancer,respectively. 4. Discussion The resultsfromtheadjustedmodelshowedthatpatients with SCIincurredasignifcantly lowerriskforsubsequent prostate

cancer,comparedwithmembersofthegeneral Taiwanese population.ThesiteoftheSCI(spinalcord level) appearedtoinfuence thepatient's riskofprostate cancer, withhigher-levelinjuriesbeingmarginallybut signifcantly associatedwithalowerrisk,whereaslower- level

injuriesappearednottoaffecttherisk.Bycontrast,the analysis

ofbladdercancerdatashowedthatpatientswith SCI werenotatincreasedriskforthiscancercomparedwith the generalTaiwanesepopulation. The datafromSurveillanceEpidemiologyandEnd Results

intheUnitedStatesshowthatoverallcancer incidence ratesforallracialandethnicgroupscombined

decreased by0.8%peryearbetween2003and2007 [21]. However,

thetrendsdifferinTaiwan,wherecancerhas been theleadingcauseofdeathinthegeneralpopulation since 1982.Theage-adjustedincidenceforcancerin Taiwan hasincreasedsteadily,reaching276newcases per 100,000peoplein2008 [7]. Cancerthusremainsa challenge forthepublichealthsystemofTaiwanandhas gained governmentattention,resultinginapopulation- based investigationoncancer-

preventionepidemiology. The NHIRDcontainshospitalservicerecords,ambulatory service

records,andprescriptionclaimdata.Itallows investigatorstoselectspecifc studygroupsandmatched comparison groupsandensuresthatsamplesarerepresen- tative

oftheunderlyingpopulationgroups.Thecurrent study usedtheNHIRDdatasourcetoinvestigatethe possible relationshipsbetweenSCIandtherisksofprostate cancer andbladdercancer. The

maindemographiccharacteristicsofpatientswith SCI inthisstudywerecompatiblewiththosereportedin similar studies.Most(62.8%)ofourpatientswithSCIwere men andwererelativelyyoung(45.8%aged49yor younger). ThemostcommonsiteofSCIwasthecervical spine, followedbythelumbarspine,andthenthoracic spine. Our fndings werecompatiblewithearlierobserva- tional

epidemiologicalstudiesthatreportedasignifcantly lower incidenceofprostatecanceramongmenwithSCI compared withage-matchedcontrolswithoutSCI [8–12]. The

reasonbehindthelowerincidenceofprostatecancer among patientswithSCIcomparedwiththegeneral population remainsunclear.Plausiblemechanismsthathave been

proposedincludeinterruptionoftheneurohormonal pathways throughextensivespinalcorddamage [8–

10,22– 24]. Theresultsofanimalexperimentshaveshownthat neurologic

andhormonalfactorsplayakeyroleinthe growth oftheprostateglandinrats [10,25,26]. However,in humans, theinfuence ofneurologicandhormonalfactors on

theprostateglandispoorlyunderstood.Severalstudies have

observedatendencytowardlowerprostatevolume and serumprostatespecifc antigen(PSA)levelaswellas testosterone defciency amongpatientswithSCI;these tendencies

mightmediatetherelevantmechanismeither directly orindirectly [9,27,28]. PriorreportsfromPannek et al. [27] and Scottetal. [29] demonstratedlowerincidence of

prostatecancerdiagnosis.Anotherstudyreportedthe clinical observationthatincidenceofprostatecanceramong patients

withmyelopathywasloweramongmoreseverely paralyzed patients [30]. Frisbie [11]

conductedaconfrma-tory studyandshowedthatpatientswithhigherspinal levels

(T10orabove)ofparalysisincurredalowerriskof prostate cancercomparedwithlessparalyzedornonpar- alyzed men.Becauseoflimitationsinourdata,wewere unable toidentifytheexactlocationofSCIasaboveor below T11.Instead,accordingtoeachpatient's ICD-9-CM diagnosis,

wecategorizedthelevelofinjuryasbelowor above T6.DespitethisestimationofSCIlocation,our analysis showedthatonlypatientswithhigher-levelinjuries were atasignifcantly lowerriskforprostatecancer.

Bartoletti etal. [9] showed thatpatientswithearly-onset SCI werelikelytohavelowerPSAlevelsandsmaller prostate glandscomparedwithpatientswithSCIwhose injuries hadoccurredlaterinlife.The fndings

ofPateletal. [8] similarly suggestedalowerincidenceofprostatecancer among

veteranswithchronicSCI.Ourresultsshowedthat the riskofoccurrenceofprostatecancercontinuedto decrease aftertheinitial5yearsofSCI,butthedifference was notstatisticallysignifcant. Several

retrospectivestudieshavesuggestedthatthe incidence ofbladdercancerinpatientswithSCIis signifcantly higherthanthatinthenormalpopulation [13–16]. However,Subramonianetal. [31] showed that the age- standardizedincidenceofinvasivebladdercancerin patients withSCIdidnotdifferstatisticallyfromthatofthe general population.Pannek [32] also foundthatthebladder cancer

incidenceinpatientswithSCIandinthegeneral population iscomparable.Our fndings wereconsistentwith those 2reports.Severalstudieshavesuggestedthatchronic irritation

inducedbyanindwellingcatheter,aswellas interactionbetweenthebladdermucosaandahighvolume of urine,asinneurogenicbladderproblems,mightbe associated withincreasedriskforbladdercancer [13–

15]. As shownin Table 2, ouranalysisrevealedthatpatients with

SCIyoungerthan50yearstendedtobeatahigherrisk for bladdercancercomparedwitholderpatients.This fnding wascompatiblewithareportbyWestetal. [33] which

showsthatbladdercancertendstobediagnosedata younger ageinpatientswithSCIcomparedwiththegeneral population. The

strengthofthisstudywasitslargesampleand population-baseddesign,whichincreasesthegeneralizabil- ity oftheresults.Thisanalysisprovidedarealisticportrayal of

theincidenceofprostateandbladdercanceramong patients withSCIinTaiwan.However,thestudywas subject tocertainlimitations.First,theNHIRDdidnot provide

informationoncertainvariablesthatwouldhave been relevanttoourinvestigation,suchasindwelling catheter, diagnosisandtreatmentofchronicinfections, stones,

histologytypeandgrade,stage,smoking,oralcohol use. Thus,wewereunabletoconductmoresophisticated tests adjustingforvariablesthatwerenotrecorded.Second, patients

withSCIwerelikelytohavelessPSAtestsfor prostate cancer.Itisalsopossiblethatmenwithsevere lesions aretestedlessthanmenwithminorlesions.The same holdsfortheobservationthatmenwhohavealonger duration ofSCIhavealowerrisk.However,thedatafor patients

withSCIwerederivedfromtheinpatientclaimsof NHIRD,whichdidnotcoverallthePSAscreening information,sowecannotanalyzeandseeifthereisany difference

inthePSAscreeningpercentagebetweenSCI cohort andnon-SCIgroup.Thesameconcernappearedin cystoscopyorcytologyforbladdercanceraswell.Apart from

thesepotentialproblems,thedataonSCIandcancer diagnosis usedinthisstudymustbeconsideredhighly reliable. In conclusion,patientswithSCIwereshowntobeata lower

riskforprostatecancercomparedwiththecompar- ison group,especiallyamongpatientswhoseSCIhad occurred inthehigherspinalcordlevel.Theriskofbladder cancer didnotdiffersignifcantly

betweentheSCIandnon- SCI cohorts.PatientswithSCImighthavelesscancer screening owingtoinconvenience,whichmightconfound the

studyresults.Furthermore,itremainsdebatablewhether these

patientsshouldbescreenedforeitherbladderor prostate cancer.